There was an issue with prepared statements and extra round trips to the DB with postgres.js. Has th
There was an issue with prepared statements and extra round trips to the DB with postgres.js. Has that been resolved?
AWe saw a reduction in latency by using the node-pg driver
MQuite some time ago - we discussed that here too. https://developers.cloudflare.com/hyperdrive/platform/changelog/#2024-05-22
AOh. Let me check again. I remember trying it after sometime in June/July and still having the issue.
Lfor hyperdrive, could I connect to it via a pages app or would I have to use worker sites?
AHyperdrive does work with Pages, yes.
Ldoes Hyperdrive have static IPs I can allow traffic from my RDS Postgres instance?
ANot yet, no. It's on our radar but for now we're still on the shared pool Warp uses, that is not static
AWhen that changes it'll warrant a changelog entry, for sure.
CHas there been any progress in securing Hyperdrive? (sorry - wrote this before seeing the post indicating there are no static IPs yet)
I know Cloudflare has deemed Hyperdrive "producton ready", but it is unusable in its current state for anything more than a sandbox... and certainly not anything targeting the enterprise. The restriction to open 0.0.0.0/0 is a widely known security vulnerability. I did this (since there was no other way), and instantly regretted it. Within a day, my Google Cloud SQL server was hammered with connections attempting to gain access... enough so that the small scale test DB instance I had was shut down for memory consumption (deemed underprovisioned... when there was literally one record getting 5 requests/day in the tiny DB). Google has plenty of ways to secure the DB, but Cloudflare is essentially requiring users to bypass these mechanisms.
I know Neon, PlanetScale, & Supabase are supported... but just supporting Guillermo Rauch-backed startups is not enough to call it "production ready". Contrary to what has been stated, most databases are NOT publicly accessible. Only managed service providers do this, and even those have the ability/recommendation to add security rules limiting connections. I mention this because it was one of the arguments made in favor of allowing 0.0.0.0/0 a few months ago in this thread.
I'd really love to use Hyperdrive. While my servers were left alone, it seemed great. However; this is a deal-breaker if it's going to put my other infrastructure in jeopardy. If there's anything I can do to help, I'd be happy to.
@AJR - is the pool of IP's for warp accessible? If we can restrict by that, it would be a start.
cc @Matt
I know Cloudflare has deemed Hyperdrive "producton ready", but it is unusable in its current state for anything more than a sandbox... and certainly not anything targeting the enterprise. The restriction to open 0.0.0.0/0 is a widely known security vulnerability. I did this (since there was no other way), and instantly regretted it. Within a day, my Google Cloud SQL server was hammered with connections attempting to gain access... enough so that the small scale test DB instance I had was shut down for memory consumption (deemed underprovisioned... when there was literally one record getting 5 requests/day in the tiny DB). Google has plenty of ways to secure the DB, but Cloudflare is essentially requiring users to bypass these mechanisms.
I know Neon, PlanetScale, & Supabase are supported... but just supporting Guillermo Rauch-backed startups is not enough to call it "production ready". Contrary to what has been stated, most databases are NOT publicly accessible. Only managed service providers do this, and even those have the ability/recommendation to add security rules limiting connections. I mention this because it was one of the arguments made in favor of allowing 0.0.0.0/0 a few months ago in this thread.
I'd really love to use Hyperdrive. While my servers were left alone, it seemed great. However; this is a deal-breaker if it's going to put my other infrastructure in jeopardy. If there's anything I can do to help, I'd be happy to.
@AJR - is the pool of IP's for warp accessible? If we can restrict by that, it would be a start.
cc @Matt
MThe pool of IPs used right now is not stable. Do not restrict by that. We’re working on a better approach here that will allow you to use Cloudflare Tunnel + Access for cases where you can’t open your firewall to the Internet.
IP ACLs aren’t/would not be scoped per customer, and we don’t think they’re the right approach.
IP ACLs aren’t/would not be scoped per customer, and we don’t think they’re the right approach.
C@Matt - Timeframe?
A@Corey this is something we're well aware of and addressing it has been our top feature priority. The Tunnel support and our move over to a stable IP pool are under active development. Matt will have to speak to timeframe/announcements etc, but I just wanted to chime in and say that we do hear you loud and clear on this one.
TTo give you an idea of timeline, we’re testing internally. So actually coming soon
TIf (and only if) you've been wanting to connect from Hyperdrive to private databases over tunnels:
@AJR and I talked it over and we're ready for you to try Hyperdrive with tunnel support for private databases early (if you're interested). Docs are still in progress, there may still be some rough edges and it's cli/api heavy (no UI yet), but we're looking forward to getting this in your hands and get your early feedback on it. If you're in,
this message we'll start a separate message thread for us to discuss this later this week.
CC @Corey @berkinovish @amcclosky (because you've brought it up in the past)
@AJR and I talked it over and we're ready for you to try Hyperdrive with tunnel support for private databases early (if you're interested). Docs are still in progress, there may still be some rough edges and it's cli/api heavy (no UI yet), but we're looking forward to getting this in your hands and get your early feedback on it. If you're in,
this message we'll start a separate message thread for us to discuss this later this week.CC @Corey @berkinovish @amcclosky (because you've brought it up in the past)
Mwill we ever be able to set caching at the query level? since hyperdrive is parsing queries it seems like you could have a statement to set cache ttl
AYes, it's on our backlog. It's not something we've really set about yet, so the exact design and behavior of it is an open question, but disabling cache and setting max-age per query are both going to happen at some point.
IThis was awesome timing with the previous message :lul:
IYou weren't kidding
AI can speak to the compatibility question, so I'm going to try to answer that one.
We're targeting MySQL. MariaDB will be handled to the extent that drivers and protocol compatibility with MySQL give it to us for free.
Explicit MariaDB support might be a future consideration but is not a current one.
We're targeting MySQL. MariaDB will be handled to the extent that drivers and protocol compatibility with MySQL give it to us for free.
Explicit MariaDB support might be a future consideration but is not a current one.
Pthe docs for the hyperdrive have some mistakes:
on create it says
vs on
also, on update, the --database is the only one who has no "origin" prefix - seems to be inconsistant, also why not just enable the same --connection-string parameter, when you anyway need to update everything
on create it says
wrangler hyperdrive create <ID> [OPTIONS]vs on
wrangler hyperdrive update <ID> [OPTIONS]also, on update, the --database is the only one who has no "origin" prefix - seems to be inconsistant, also why not just enable the same --connection-string parameter, when you anyway need to update everything
NHey team, is it possible to use Hyperdrive directly from Pages rather than through Workers? Or are there plans for this in the future?
For context, we are connecting to our
But we were wondering if we could also leverage Hyperdrive.
For context, we are connecting to our
pgBut we were wondering if we could also leverage Hyperdrive.
KHi Cloudflare team, we’ve been using hyperdrive for a while now and we really like it. Are there any plans on making hyperdrive available outside of workers?
Pa little bit more complicated but what i would wish:
that the localConnectionString would move out of the toml, maybe into the .dev.vars file - that i can give it a key. on our system it happens that the local strings get committed to the sourcesafe. not great.
that the localConnectionString would move out of the toml, maybe into the .dev.vars file - that i can give it a key. on our system it happens that the local strings get committed to the sourcesafe. not great.
KRecommended Create aenvironmental variable with the connection string of your database. <BINDING_NAME> is the name of the binding assigned to your Hyperdrive in your wrangler.toml or Pages configuration. This allows you to avoid committing potentially sensitive credentials to source control in your wrangler.toml, if your test/development database is not ephemeral. If you have configured multiple Hyperdrive bindings, replace <BINDING_NAME> with the unique binding name for each.WRANGLER_HYPERDRIVE_LOCAL_CONNECTION_STRING_<BINDING_NAME>
CThanks - I already implemented an alternative solution.
JMongoDB with Hyperdrive?
Is their a difference between:
connecting a MongoDB directly
connecting a MongoDB with Hyperdrive
Is their a difference between:
connecting a MongoDB directly
connecting a MongoDB with Hyperdrive
SMongoDB is not supported with Hyperdrive atm
TConfirming what Sam said
LHyperdrive can be used from one aws lambda or only inside a worker?
AOnly from within a Worker.
A@hyperdrive
For all of you who've asked about connecting to your database without exposing it to the public Internet, we're ready to go live. Say hello to Hyperdrive, now with support for Cloudflare Tunnels!
https://developers.cloudflare.com/hyperdrive/configuration/connect-to-private-database/
For all of you who've asked about connecting to your database without exposing it to the public Internet, we're ready to go live. Say hello to Hyperdrive, now with support for Cloudflare Tunnels!
https://developers.cloudflare.com/hyperdrive/configuration/connect-to-private-database/
Cloudflare Docs
Hyperdrive can securely connect to your private databases using Cloudflare Tunnel and Cloudflare Access.
Sletsgo!
Mdamn!
M@Hyperdrive new feature just dropped!
C
AThanks Matt. Pings are hard, apparently....
HSo if I am reading this correctly, Hyperdrive needs a public domain for the tunnel, correct? You can't point Hyperdrive directly at the ArgoTunnel domain(I don't remember the exact domain name off the top of my head)
ACorrect.
HWould it be possible to make Tunnels/Hyperdrives on the same account be able to communicate with each other directly?
AIn what sense? The traffic flow right now is
Worker<->Hyperdrive<->DatabaseLike, where I give my Hyperdrive a Tunnel ID, then it connects directly, so that there is no outside evidence that I have a connection to my DB exposed
AAh. We're poking at that, it's a possible future option. CC @thomasgauvin
PHow does this impact performance?
AIn almost all scenarios it's going to be very comparable to current performance. I wouldn't reach for this to solve perf problems, but I wouldn't expect it to be slower than your current setup.
AWith (really) big bulk inserts you might see some slowdowns. Please let us know if you do, as that's something we would want to address.
PI don't personally. Only case would be data import but there's an inbuilt 4 second budget for that and if that budget is surpassed there's no real consequences
AI meant if you do try it and do see slowdowns. We know some folks do larger inserts, but we don't think it should move the needle much for those. It's just an area of interest.
IOk
